This presentation, from Dr Sarah Crozier, focused on research exploring how the association between an unhealthy childhood diet and body composition depends on prenatal experience. The developmental mismatch hypothesis proposes that risk of diseases such as obesity is increased when impaired prenatal nutrition and growth, is followed by an unhealthy childhood diet. This project used data from the Southampton Women’s Survey (SWS) to investigate whether there was an interaction between conditional growth in fetal abdominal circumference (AC) in late pregnancy and diet at age 6 years on body composition at age 9 years.
SBFT Tool Competition 2024 -- Python Test Case Generation Track
The association between an unhealthy childhood diet and body composition depends on prenatal experience
1. The association between an unhealthy
childhood diet and body composition
depends on prenatal experience
Sarah Crozier, Janis Baird, Hazel Inskip, Nick Harvey, Sian
Robinson, Cyrus Cooper, Mark Hanson and Keith Godfrey
Medical Research Council Lifecourse Epidemiology Unit, Southampton, UK
2. Introduction
• The prevalence of obesity is rising in children and is
associated with both childhood ill health and an
increased risk of subsequent adult obesity
• Intrauterine life may be a critical period for the
programming of later obesity
• The developmental mismatch
hypothesis proposes that risk of
diseases such as obesity is increased
when impaired prenatal nutrition
and growth, is followed by an
unhealthy childhood diet.
3. Southampton Women’s Survey
12,583 non-pregnant Southampton women aged 20-34,
interviewed about diet, physical activity,
social circumstances and lifestyle.
3,158 live-born singleton births.
Offspring followed through
pregnancy, infancy and beyond.
6. Food frequency questionnaire
• The broad pattern of 6 year diet has been
characterised by the use of a prudent diet score.
7. Principal component analysis
Weighting Frequency
(per week)
Total
White bread
Crisps
-0.20 0.3 = -0.1
-0.21 × 0.5 = -0.1
Total = 3.2
×
Green vegetables
Salad vegetables
0.33 7 = 2.3
0.25 4.5 = 1.1
×
×
8. Median frequency per week food intake by quarters of
the 6 year prudent diet score
Food Least prudent
quarter
Most prudent
quarter
Salad vegetables 0.8 6
Green vegetables 2.3 6.5
Root vegetables 2 4.3
Other vegetables 0.5 2
Crisps 5 2
Processed meat 7 4.8
White bread 7 1
Chips and roast potatoes 3 1.5
9. Outcomes
• Dual-energy X-ray
Absorptiometry
(DXA) was used to assess body
composition at 9 years; fat, lean
and bone mass were derived
using paediatric software.
• 592 children included in the
analysis.
10. Directed Acyclic Graphs
• An analysis can stand or fall on the choice of
confounders
• A Directed Acyclic Graph (DAG) or causal diagram
describes a model of the associations between all
variables that could influence the exposure-
outcome association.
11. DAG
• Adjust for: 9 year height, sex, breastfeeding duration,
maternal BMI, education, smoking in pregnancy, late
pregnancy vitamin D and pregnancy weight gain
Maternal
education
Pre-
pregnancy
BMI
9 year
height
Late
pregnancy
vitamin D
Sex
9 year body
compositio0n
AC growth
Smoking in
pregnancy
Age at DXA
Pregnancy
weight gain
Duration of
breastfeeding
12. Characteristics
Characteristic
Maternal education ≥ A-levels, n (%) 376 (63.6%)
Pre-conception BMI, kg/m2 [median (IQR)] 24.0 (22.1 to 27.0)
Female, n (%) 303 (48.8%)
Age at DXA scan, years [mean (SD)] 9.2 (0.3)
9 year total fat, kg [median (IQR)] 7.6 (5.7, 10.0)
9 year total lean, kg [mean (SD)] 22.7 (3.3)
9 year total BMC, kg [mean (SD)] 1.0 (0.1)
n = 592
13. Main effects - fat mass
n = 199
-.2
0
.2
.4
.6
.8
Regressioncoefficient(95%CI)11
w
eek
size
11-19
w
eek
grow
th
19-34
w
eek
grow
th
34
w
eek-birth
grow
thBirth-6
m
onth
grow
th6-12
m
onth
grow
th
12
m
onth-2
yeargrow
th
2-3
yeargrow
th
3-6
yeargrow
th
9 year total fat (SDs)
14. Main effects – percentage fat
n = 199
-.2
0
.2
.4
.6
.8
Regressioncoefficient(95%CI)11
w
eek
size
11-19
w
eek
grow
th
19-34
w
eek
grow
th
34
w
eek-birth
grow
thBirth-6
m
onth
grow
th6-12
m
onth
grow
th
12
m
onth-2
yeargrow
th
2-3
yeargrow
th
3-6
yeargrow
th
9 year percentage fat (SDs)
15. Main effects – total lean
n = 199
-.2
0
.2
.4
.6
.8
Regressioncoefficient(95%CI)11
w
eek
size
11-19
w
eek
grow
th
19-34
w
eek
grow
th
34
w
eek-birth
grow
thBirth-6
m
onth
grow
th6-12
m
onth
grow
th
12
m
onth-2
yeargrow
th
2-3
yeargrow
th
3-6
yeargrow
th
9 year total lean (SDs)
16. Main effects – total BMC
n = 199
-.2
0
.2
.4
.6
.8
Regressioncoefficient(95%CI)11
w
eek
size
11-19
w
eek
grow
th
19-34
w
eek
grow
th
34
w
eek-birth
grow
thBirth-6
m
onth
grow
th6-12
m
onth
grow
th
12
m
onth-2
yeargrow
th
2-3
yeargrow
th
3-6
yeargrow
th
9 year total BMC (SDs)
17. Interaction – total fat
P-value for
interaction = 0.006
-.4
-.2
0
.2
.4
9yeartotalfat(SDs)
-0.7 -0.2 0.2 0.7
Conditional AC growth 34 weeks to birth (SDs)
-0.7 -0.2 0.2 0.7
Prudent diet score (SDs)
19. Interaction – total lean
P-value for
interaction = 0.97
-.2
-.1
0
.1
.2
.3
9yeartotallean(SDs)
-0.7 -0.2 0.2 0.7
Conditional AC growth 34 weeks to birth (SDs)
-0.7 -0.2 0.2 0.7
Prudent diet score (SDs)
20. Interaction – total BMC
P-value for
interaction = 0.94
-.3
-.2
-.1
0
.1
.2
.3
9yeartotalBMC(SDs)
-0.7 -0.2 0.2 0.7
Conditional AC growth 34 weeks to birth (SDs)
-0.7 -0.2 0.2 0.7
Prudent diet score (SDs)
21. Strengths and weaknesses
• Strengths
Detailed anthropometric measurement and conditional
growth analyses enabled description of abdominal
circumference growth.
Dietary patterns describe broad patterns of diet, with
greater potential for public health intervention.
Directed acyclic graphs provide an objective method to
determine confounders, aiming to describe causality.
• Weakness
The conditional growth method only provides measures of
fetal growth for participants with abdominal circumference
data at all time points.
22. Conclusions
• Individuals showing late gestation faltering of fetal
growth who then had an unhealthy childhood diet had
greater adiposity, while childhood diet was less
influential on adiposity in individuals whose fetal
growth had not faltered.
• There were no similar interactions for lean and BMC
outcomes.
• The result for adiposity provides some evidence in
support of the mismatch hypothesis.
23. Acknowledgements
Funding:
• Medical Research Council
• University of Southampton
• Dunhill Medical Trust
• British Heart Foundation
Ultrasonographers
Doctors
Administrative staff
Nurses and Midwives
Telephonists
12,583 SWS study participants
Dieticians/nutritionists
Research assistants
Clerical staff
Laboratory staff
Statisticians
Computing staff
• Food Standards Agency
• National Institute for Health
Research
• European Union’s Seventh
Framework Programme